A Tel Aviv college has teamed up with Italian researchers to build an app that will tell people whether they are likely to have coronavirus, based on their voice and breathing.
“Our cellphone app will ask you to say a few sentences and cough for a minute or two, compare your voice with templates it has for coronavirus-positive and -negative [patients], and tell you the chances that you are infected,” said Ami Moyal, president of Afeka Tel Aviv Academic College of Engineering.
Many doctors have noted that coronavirus can have an impact on patients’ speech and on their breathing patterns, often causing them to catch their breath while talking.
Currently, many people agonize over whether they should be getting tested, with much confusion about symptoms and how they are experienced. It can be an especially hard decision, as registering for a test can have major consequences on lifestyle. In many countries, including Israel, people are instructed to self-isolate from the moment they request a test at least until they receive a result, which can be several days.
Afeka’s app is intended to take the mystery factor out of the equation, by telling people the probability that they have coronavirus, enabling them to make a clear-headed decision on whether to go for screening.
It will also help to spot some carriers while they are still asymptomatic, potentially detecting their illness before they infect others, said Moyal, commenting: “The idea is to target patients in the first few days while they don’t have have symptoms.”
He added that as the app, which is expected to take a few months to complete, will just listen to the voices of users, not register what they are saying. It will be suitable for people speaking any language.
Some experts say that, even if a vaccine is developed, coronavirus will remain a problem for years, and there are questions as to whether swab testing labs will be able to keep up with demand. This has driven interest in pre-screening tools like Afeka’s.
Italy’s ambassador to Israel, Gianluigi Benedetti, visited Afeka this week and made an agreement to commit the Catholic University of the Sacred Heart in Milan to take part in the research and jointly develop the app. The University of Cambridge will contribute analysis of breath and coughs needed for the project. All participants have submitted a joint request for funding to the European Union for a fund for fast-tracked coronavirus projects.
Afeka started the project early in the pandemic, when the Defense Ministry launched a program to encourage coronavirus-related innovation. Now it is working with its new partner to record audio from around 1,000 healthy people and 1,000 people with coronavirus, in order “teach” the app that is being produced what sounds to look out for to warn of possible infection.
They are also gathering audio from people with flu, so that the app will know not to throw up false-positive results for people who have regular influenza.
“Test protocols are predefined by the Afeka Center for Language Processing, together with leading physicians, laryngologists and heads of coronavirus wards,” said Moyal. “These recordings will be analyzed by specially-developed machine learning systems that extract characteristics unique to the vocalizations of COVID-19 positive patients.
“The system will later be trained to recognize these characteristics in new speakers. At the end of the process, there will be a cellphone application that is used to record and indicate whether a user is suspected of having contracted COVID-19.”
During his visit, Benedetti paid tribute to “long-term intensive contacts between the Italian and Israeli scientific communities, further strengthened since the very first stage of the pandemic.”